Research On Unknown Node Positioning In WSN Based On Fruit Fly Visual Neural Network And DV-Ho

Wireless sensor network,which can perceive and send sensor information in real time,is a wireless self-organization network with static and mobile sensors.The functions of its location information acquisition and tracking mobile or unknown nodes have been comprehensively applied to many fields,e.g.environmental monitoring,public safety,urban intelligent transportation and so on.However,although some location-based positioning methods have played an important role in the research of node positioning,it is still open how to design node positioning models and efficient and universal positioning algorithms in the fields of science and technology.Therefore,based on the classical DV-Hop positioning model,the thesis develops unknown node localization optimization models in static,noisy,and dynamic environments as well as related hybrid visual evolution neural networks.Herein,the neural networks are constructed based on the inspiration of population evolution and the fly and locust’s information processing mechanisms.The acquired achievements can not only help to promote the joint development of intelligent optimization and computer vision,but also provide an important reference for solving the problem of unknown node location in WSN.The main work and achievements are summarized below:A.An improved DV-Hop positioning model is developed to avoid the shortcomings of the low hop distance estimation accuracy of the DV-Hop positioning model in static environment and the difficulty of accurately locating the unknown nodes’ positions in WSN.Meanwhile,a hybrid visual evolutionary neural network is designed to solve such a model.In the design of the model,four kinds of communication radiuses are introduced to calculate the number of hops between nodes,while a dynamic weight factor is employed to rectify the minimum hop number and the average hop distance calculation model in order to obtain an improved DV-Hop calculation model.Subsequently,a hybrid visual evolutionary neural network is designed to tackle the DV-Hop localization problem.Herein,a hybrid visual neural network is constructed to output activities called global or local learning rates by means of the information processing mechanisms of fly and locust visual systems,in which each candidate solution and the values of the objective function are viewed as a state and gray-scale values,respectively.Additionally,a state matrix,which bases on the harris hawks’ position update strategy,is constructed to transfer the current states toward potential regions.Numerically comparative experiments show that the visual evolutionary neural network has significant advantages in solving benchmark function optimization problems and can effectively locate unknown nodes.B.Related to the factors of which the noise environment seriously disturbs the unknown nodes’ positioning in WSN and degrades the accuracy of DV-Hop positioning,the unknown node positioning problem,which overly depends on hop nodes’ distance estimation,is transformed into an expected value planning problem.Subsequently,a hybrid visual evolutionary neural network is proposed to seek the positions of the unknown nodes in WSN.Therein,a hybrid visual neural network is constructed to output activities called global and local learning rates in terms of the information-processing mechanisms of the fly and locust’s visual systems,in which the learning rates are used to guide the states’ transition in terms of a state update strategy.Theoretical analysis shows that the computational efficiency of the evolutionary neural network is determined by the sample size of random variables,the input resolution of the visual neural network,and the dimension of the optimization problem.Comparative experiments show that the neural network with simple structure and applicability can effectively solve unknown node positioning problems with a high positioning accuracy and the abilities of strong noise suppression and fast solution search.C.Aiming at the problem of the unknown nodes’ positioning in dynamic environments,a dynamic positioning optimization model of unknown nodes is here designed,where the average positioning error is taken as the performance index.Combined to the real-time requirement of unknown node positioning,the environmental change detection scheme is designed to detect whether a time-dependent environment makes a change.By appropriately improving the convergence layer’s computing model in the existing visual neural network,an improved hybrid visual neural network is designed to output activities named global or local learning rates.Then,related to the particle update strategy of the basic swarm optimization approach,a state update scheme is combined with the visual neural network to obtain an improved hybrid visual evolutionary neural network.Comparative experiments show that the visual evolutionary neural network can satisfy the positioning requirement of dynamic unknown nodes with stable solution search effect and strong environment tracking ability.